Feels like a variant of the “prompt injection” vulnerabilities that are so amusing in LLMs. Call it “knowledge graph injection”. Perhaps it’s a fitting penalty for the seemingly common practice of making support numbers difficult to find (or worse, nonexistent)
> Google is one of the prominent and top-rated company in all over the world because it offers the best class multiple products and services to their users.
I fear that ChatGPT will greatly improve Scammer English.
Here’s a screenshot in case it changes or is different for others: https://ibb.co/qkvqKjW